
EU AI Act Article 50
EU AI Act transparency provision requiring machine-readable marking of AI-generated synthetic content; in-market systems grandfathered to 2 December 2026.
Last refreshed: 15 July 2026 · Appears in 1 active topic
Do AI tools already in use by broadcasters still face the 2 August Article 50 deadline?
Timeline for EU AI Act Article 50
Two EU clocks strike on 22 July
Media's AI PivotBrussels stalls its own AI-label code
Media's AI PivotMentioned in: Paramount-WBD deal stalls at three gates
Media's AI PivotMentioned in: WPP convenes six rivals on AI buying
Media's AI PivotSpotify adopts DDEX to label AI tracks
Media's AI PivotWhat does EU AI Act Article 50 require broadcasters to do?
When does EU AI Act Article 50 come into force?
Does EU AI Act Article 50 apply to AI-generated news text?
Background
Article 50 of the EU AI Act establishes transparency obligations for providers and deployers of AI systems generating synthetic audio, image, video, or text. The original enforcement date was 2 August 2026. Following the AI Omnibus provisional agreement of 7 May 2026, AI systems already on the market before 2 August 2026 were grandfathered, giving them until 2 December 2026 to meet the machine-readable marking requirement. Any new AI deployment launched after 2 August faces the full Article 50 standard from the moment it goes live with no grace period. The Code of Practice operationalising Article 50 was finalised on 10 June 2026 following the closing plenary. Providers and deployers who want the presumption of conformity must submit a completed signature form by 22 July 2026, 18:00 CEST, to be published on the initial-signatory list before the Act's 2 August general application date; non-signatories face a heavier evidentiary burden proving an alternative approach is equally effective. As of 15 July 2026, no EU broadcaster or media company has publicly signed. Under the Code, deployers must label AI-generated text on matters of public interest, unless that text sits under human editorial responsibility.
The technical requirement is specific: providers must ensure AI-generated outputs carry a machine-readable format detectable as artificially generated or manipulated. Deployers of deepfake systems must disclose that content has been artificially generated; publishers of AI-generated text on matters of public interest must disclose the AI origin. Article 50 is distinct from the risk-tier framework in the parent EU AI Act, and from the separate GPAI Code of Practice, which governs frontier-model provider obligations rather than content marking. Spotify's adoption of the DDEX standard at its Investor Day on 21 May 2026 illustrated the compliant-by-design route: threading AI-provenance flags through existing licensing-metadata pipes rather than retrofitting a separate marking layer under deadline pressure.
The Article 50 marking obligation applies to any service reaching EU audiences regardless of where the provider is headquartered, meaning US-based platforms including YouTube, Apple Music, and TikTok face the same compliance calendar as European broadcasters. The 22 July signatory cutoff lands the same day the Commission's separate Paramount-Warner Bros. Discovery merger-control decision is now due, an unrelated coincidence of timing rather than a linked process. France Télévisions' vendor-disclosure approach at Roland-Garros on 11 June 2026, naming Decart, Gracia AI, and Topaz Labs as its live-production AI tools, is emerging as a transparency-forward template for how large European public broadcasters operationalise Article 50.